MCP server for searching and interacting with 1C Enterprise 8 documentation using RAG (Retrieval-Augmented Generation) with embeddings and Qdrant vector database
The server exposes a single tool 'search_1c_documentation' with a well-structured schema and reasonable descriptions. However, the implementation is incomplete: the tool definition is visible in the schema specification but critical implementation details are absent from the provided source code. The schema includes proper typing and enums for object_type filtering, which is good. Descriptions are present but relatively generic and could be more action-oriented for LLM selection. The server appears to be in early development (evident from incomplete Dockerfile and missing embedding_service.py implementation). Since this is a fastmcp HTTP server, protocol transport is correct, but the lack of visible tool implementation code and missing error handling patterns limit the overall quality score.
Поиск описания объектов конфигурации 1С Предприятие 8 в документации.
Tool implementation not visible in provided source code. Only schema declaration is evident; actual search logic, error handling, and response formatting are absent or in unshown files.
Tool description lacks clarity on WHEN to use this vs. other discovery tools and what exactly is returned. Current description: 'Поиск описания объектов конфигурации 1С Предприятие 8 в документации.' is minimal (64 chars in Russian). Should explain: returns structured documentation excerpts, limited to 1C v8 docs, and what fields are in the response.
No documented output schema. The rubric requires tools to document what fields they return so LLMs know what to expect and can chain downstream calls. Returns are not specified in visible code.
Inferred effective spec: 2025-06-18+.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-22 | C | 62 | 2025-06-18+ | v2 |
| 2026-03-09 | F | 46 | - | v1 |
No error handling guidance. What happens if query is empty? If no results match? If the 1C documentation service is down? Error responses should guide recovery: e.g., 'No results for "ДокументОснование". Try a shorter search term or check the object_type filter.'
Parameter 'limit' lacks minimum/maximum constraints. Unbounded numeric parameters invite abuse (e.g., limit=999999). Should specify 1 - 100 or similar in description and schema.
The 'query' parameter description is in Russian and does not explain what happens if the query is empty, too short, or contains special characters. LLMs need clear guidance on valid input ranges.
No pagination or result limit enforcement documented. If the tool can return hundreds of matches, the description should state: 'Limited to X results; use query refinement for more specific results.' The baseline rubric requires enforcing result limits to prevent context window exhaustion.